SSSE3 high bit depth sad_aligned functions

diff --git a/common/pixel.c b/common/pixel.c
index 32809a2c2068926d2685f221d3994d04d4feb085..8c3c196348593213a751dff8633d20a6a64919c8 100644 (file)
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -825,6 +825,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     }
     if( cpu&X264_CPU_SSSE3 )
     {
+        INIT4_NAME( sad_aligned, sad, _ssse3_aligned );
         INIT7( sad, _ssse3 );
         INIT7( sad_x3, _ssse3 );
         INIT7( sad_x4, _ssse3 );

diff --git a/common/x86/pixel.h b/common/x86/pixel.h
index dfedd7a5a1b3ae50267b1f53b328fa2d36fb1416..50c9faf1834b96a0fe83d49e69be06d6182deb46 100644 (file)
--- a/common/x86/pixel.h
+++ b/common/x86/pixel.h
@@ -50,6 +50,7 @@ DECL_X4( sad, sse2_misalign )
 DECL_X1( sad, sse3 )
 DECL_X1( sad, sse2_aligned )
 DECL_X1( sad, ssse3 )
+DECL_X1( sad, ssse3_aligned )
 DECL_X4( sad, mmxext )
 DECL_X4( sad, sse2 )
 DECL_X4( sad, sse3 )

diff --git a/common/x86/sad16-a.asm b/common/x86/sad16-a.asm
index d19d05b389e8a222a977e270db2c37c9fdd181f8..bb480ce8c1f5a2d123caf1f39c738634cc23c5e6 100644 (file)
--- a/common/x86/sad16-a.asm
+++ b/common/x86/sad16-a.asm
@@ -189,6 +189,12 @@ SAD_XMM 16,  8, ssse3
 SAD_XMM  8, 16, ssse3
 SAD_XMM  8,  8, ssse3
 SAD_XMM  8,  4, ssse3
+%define movdqu movdqa
+SAD_XMM 16, 16, ssse3_aligned
+SAD_XMM 16,  8, ssse3_aligned
+SAD_XMM  8, 16, ssse3_aligned
+SAD_XMM  8,  8, ssse3_aligned
+%undef movdqu
 %undef ABS1
 %undef ABS2